[Sammelthread] Intel DDR5 RAM OC Thread

[perlibu]

genau den im link von Wera

ehm wo? kannst du den posten bitte?
danke

 

[Felix the Cat]

ehm wo? kannst du den posten bitte?
danke

Du zitierst den Post mit den Links und fragst, welchen er hat und fragst dann noch nach den Links?

 

[perlibu]

Du zitierst den Post mit den Links und fragst, welchen er hat und fragst dann noch nach den Links?

lol. ich hatte die Links mit den Tools zittiert. Aber ich ich hatte vorher schon gesagt ich zahl keine 200 für ein Kühler, also war ich beim Kühler. egal...

 

[AndreasP1981]

Dein Text zu beginn:

ja ich zahl keine 200 für einen direct die Kühler. oder eben so ein drehmomentschlüssel. wieviel nm muss der angezogen werden? 0.6nm ?

Meine Antwort dazu:

0,45Nm hab ich meinen ab 0,6 geht der MC SP runter 0,4 und drunter wird temp schlechter

Beitrag automatisch zusammengeführt: 10.01.2024

so schlimm kommts preislich aber nicht:

https://www.amazon.de/dp/B003KN3G96?ref=ppx_yo2ov_dt_b_product_details&th=1

https://www.amazon.de/dp/B0BD4MGRHR?ref=ppx_yo2ov_dt_b_product_details&th=1

Die war darauf bezogen das du mit einem Drehmomentschraubendreher die 200€ nicht erreichst

Beitrag automatisch zusammengeführt: 11.01.2024

Mal ein kleines update...
da ich mir am 8000er XMP Profil die zähne ausgebissen habe, habe ich vor ~einer Woche mit dem 7600er angefangen.
(ja ging ne weile, hab aber auch nicht täglich zeit gehabt, war beruflich ziemlich eingespannt...)
Durch viel lesen, rumprobieren und sehr oft ärgern hab ich wohl eine beim mir funktionierende Einstellung gefunden für 7600MT's:
Anhang anzeigen 957734

Mit abstand das schönste was ich in den letzten Tagen gelesen habe:
Anhang anzeigen 957735

Aktueller Karhu zwischenstand:

 

[AndreasP1981]

Karhu passt auch

 

[tibcsi0407]

@Veii
Hi Veii,

Do you have any idea where to start to fix the issue?
I know if I will start it by myself, I will just mess up the profile.

TX: 1.30V
SA: 1.18V
MC_VDD: 1.475V
DRAM VDD: 1.47V
DRAM VDDQ: 1.45V
VTT: 1.1V

CPU should be stable. Passed 2x 90 min on 8400C36 with the same setting.
CPU setting is P59x3 P58x5 P57x8 + 2TVB E45X R45-50X LLC4 IA_ACLL 0.22 ICCMAX 420A
Skews and training settings are default.

Hope you have an idea.
Thanks in advance!

 

[Veii]

Hi Veii,

Do you have any idea where to start to fix the issue?
I know if I will start it by myself, I will just mess up the profile.

Hello~

CPU should be stable. Passed 2x 90 min on 8400C36 with the same setting.
CPU setting is P59x3 P58x5 P57x8 + 2TVB E45X R45-50X LLC4 IA_ACLL 0.22 ICCMAX 420A

I'm really not sure on that.
Very much not.

1.43 VID requests are guaranteed throttle
That ontop of ICCMAX being "low" normal , can mean it will hit it even faster
And that then will mean that it throttles even stronger and might actually lack voltage - even if you basically overvolt now.
PL4 (ICCMAX) also counts for DDR not only the cores

WTRS is a bit high for my taste but i think all is ok
RDWR SG/DG gives buffer. Same for _DR & _DD give buffer.

While i dont like some timings,
If TM5 says its ok, it probably is

Any new updates on the Bios config
Anything i can inspect as of now ?

Beitrag automatisch zusammengeführt: 12.01.2024

VTT means Termination PLL , or which PLL ?
Any voltage offsets used besides TVB & IA_AC ??
What is VR_MAX on ?
Did you use anything for CTL0 Skews ?

EDIT:
PMIC stability looks good and perf data looks alright (if you read last like 4-5 big posts)
But those VID requests are too high for my taste

 

[tibcsi0407]

Hello~

I'm really not sure about that.
Very much not.

1.43 VID requests are guaranteed throttle
That on top of ICCMAX being "low" normal , can mean it will hit it even faster
And that then will mean that it throttles even stronger and might actually lack voltage - even if you basically overvolt now.
PL4 (ICCMAX) also counts for DDR not only the cores

WTRS is a bit high for my taste but I think everything is ok
RDWR SG/DG gives buffer. Same for _DR & _DD give buffer.

While I don't like some timings,
If TM5 says its ok, it probably is

Any new updates on the Bios config
Anything I can inspect as of now?

Beitrag automatisch zusammengeführt: 12.01.2024

VTT means Termination PLL , or which PLL ?
Any voltage offsets used besides TVB & IA_AC ??
What is VR_MAX on?
Did you use anything for CTL0 Skews?

EDIT:
PMIC stability looks good and perf data looks alright (if you read last like 4-5 big posts)
But those VID requests are too high for my taste

Thank you!
Yes, lower WR SG/DG made some trouble for me in the past, so it's a bit loose. Do you have suggestion for that? I would retest it if that could help a liittle.
VTT is the PLL Termination Voltage. Got the top from IMFodor on OCN. With that I was able to pass 20 Cycles of Usmus with a really low voltage. I needed 1.50V/1.47V (38-51-51) before, now it passed with 1.47V/1.45V (38-49-49).
I tried to keep my previous VDDQ - TX delta from my 8400C36 profile, looks like 0.15V is good for me.
My VR_MAX is 1720 mV I believe.
I don't have Vcore Voltage offsets right now at all.
Here my CPU info

CTL0 skews are auto, didn't touch any skews yet. I had some experiments in the past, but no success yet.

MC SP is 84 now, these are old pics.

I will set offsets on V/F curve like @zebra_hun did, but I am really busy nowadays and that is very time consuming. Tought maybe it's okay if it's already passed several times.

 

[Veii]

Thank you!
Yes, lower WR SG/DG made some trouble for me in the past, so it's a bit loose. Do you have suggestion for that? I would retest it if that could help a liittle.

its where it has to be ~ 22

With that I was able to pass 20 Cycles of Usmus with a really low voltage.

Its not really "low" per-se.
PLL should be not be used outside XOC scenarios.
We both know the scaling but like , its too early.
You crash at couple minutes, so how are we supposed to reach 4-6 hours

I tried to keep my previous VDDQ - TX delta from my 8400C36 profile, looks like 0.15V is good for me.

PLL will mess with it, Slopes and Groups will mess with it
VDDQ Training will mess with it
Biosupdates will mess with it
And VDD/Q delta on mem itself will mess with it
Oh right, SA will mess with it, due to side-issue of ODT messing with it.

Outside those little things, ya its fine 🤭

I will set offsets on V/F curve like @zebra_hun did, but I am really busy nowadays and that is very time consuming. Tought maybe it's okay if it's already passed several times.

Soo probably start with

P1  ~ (-69)
P2  ~ (-66)
P3  ~ (-24)
P4  ~ (+12)
P5  ~ (+6)
P6  ~ (-60)
P7  ~ (-138)
P8  ~ (-142)
P9  ~ (-146)
P10 ~ (-147)
P11 ~ (-150)

Something along those lines

My VR_MAX is 1720 mV I believe.

Down to 1550 for now
IA AC down to 0.35 for now

Then either give me a phone recording or capture somehow
SFT VST VT3 with HWInfo
Nothing else open.

Beitrag automatisch zusammengeführt: 12.01.2024

You may also want to show it visually (before opening anything else)
on ASUS OC_TOOL

^ looks solid
You understood

Your example

LLC4 IA_ACLL 0.22

Your problem arises, because IA supply factors 4 things in. At least
Its not only the the p-ccores, not only the e-cores , not only ring.
Real IMC voltage is not VDDIO.

 

[tibcsi0407]

Anhang anzeigen 958525
its where it has to be ~ 22

Its not really "low" per-se.
PLL should be not be used outside XOC scenarios.
We both know the scaling but like , its too early.
You crash at couple minutes, so how are we supposed to reach 4-6 hours


PLL will mess with it, Slopes and Groups will mess with it
VDDQ Training will mess with it
Biosupdates will mess with it
And VDD/Q delta on mem itself will mess with it
Oh right, SA will mess with it, due to side-issue of ODT messing with it.

Outside those little things, ya its fine 🤭

Anhang anzeigen 958527
Soo probably start with

P1  ~ (-69)
P2  ~ (-66)
P3  ~ (-24)
P4  ~ (+12)
P5  ~ (+6)
P6  ~ (-60)
P7  ~ (-138)
P8  ~ (-142)
P9  ~ (-146)
P10 ~ (-147)
P11 ~ (-150)

Something along those lines

Down to 1550 for now
IA AC down to 0.35 for now

Then either give me a phone recording or capture somehow
SFT VST VT3 with HWInfo
Nothing else open.

Beitrag automatisch zusammengeführt: 12.01.2024

You may also want to show it visually (before opening anything else)
on ASUS OC_TOOL

Your exampleYour problem arises, because IA supply factors 4 things in. At least
Its not only the the p-ccores, not only the e-cores , not only ring.
Real IMC voltage is not VDDIO.

Thank you, I will check it and report back. Should I keep LLC4 or set LLC 3 instead?

 

[Veii]

Should I keep LLC4 or set LLC 3 instead?

Ive factored that in

 

[robonz]

@Veii
Hi Veii,

Do you have any idea where to start to fix the issue?
I know if I will start it by myself, I will just mess up the profile.

TX: 1.30V
SA: 1.18V
MC_VDD: 1.475V
DRAM VDD: 1.47V
DRAM VDDQ: 1.45V
VTT: 1.1V


Anhang anzeigen 958516

Why would you only test ~31GB in VST/VT3 when you have 48GB in total?

 

[Veii]

I searched a lot of results for Heaven Benchmark. ~40-50 fps is the minimum for everyone.

Ya but no
It differs up to CPU and GPU Manufacture
Around 50 is common, but 51 vs 55 is a difference for example.
And on bad GPU curve also can drop to 22-25 FPS

Why would you only test ~31GB in VST/VT3 when you have 48GB in total?

Hmmm
Moin

I searched a lot of results for Heaven Benchmark. ~40-50 fps is the minimum for everyone.
Since there are 26 runs, there are 25 switches and fast charging, which is why the min fps is so low, since it also measures those.

All is on purpose
I need a weak CPU focused Benchmark which may or may not fit in cache
That runs CPU Clock dynamic
Which seems for you doesnt happen. Either due to powerplan or some other strangeness

Revi OS + Park Controll.

Didnt you say SpectreOS at first ? I'm on Revi the whole time and don't feel like updating // (this image is ~1 year old, and gets maintained)

It doesn't behave the way you described. Neither for Intel or AMD.

EDIT:
Wenn du alles schließt, auf wie viele Processes kommst du ?

88 wäre bei mir mit 28 Browser Fenster, Discord und das Screenshot Programm.

@zebra_hun
I dont want to say it, but are you confident in your powerplan behavior ?

We do want the CPU to assign load, and not be limited by userchanges (limiting jumping behavior can be an issue)
or limited by e vs p-cores recognition.
We want it as natural behaving as possible.

Revi tunes the powerplans and the scheduler, if it detects you have an AMD or Intel SKU.
But i wouldnt use those powerplans.
Let it behave balanced and natural
Soo on games that are supported, Intel APO functions and the rest is offloaded correctly.

Not every load needs to stay on p-cores
Not every background service or device focused needs to stay on e-cores.
Priority scheduling is functional and should stay that way. Not be influenced by 3rd party apps.
This is a positive of Win11.
Scheduling is good, but usually comes with an obsessive amount of bloat + analytics. Active and hidden.

 

[zebra_hun]

Guten Morgen!

Nein, ich benutze Revi OS Power Plan nicht. Nennt Ultra Perf.
Balanced ist besser. Damals habe ich das benutzt, aber Ultra bleibt standig max Takt.
W11 aber normal, wo ich immer getested. W10 ist Revi OS, aber ich werde beide OS neu installiert.
Jetzt aber beide OS funtioniert gut, gehen runter sofort 800MHz wenn idle ist.
W11 geht normal, nur W10 muss man Park Control benutzen. TM5, Cb und viele App wollen nur e Kerne nutzen. Aber mit Balanced geht gut.

 

[Veii]

Guten Morgen!

Guten Morgen

W11 aber normal, wo ich immer getested. W10 ist Revi OS, aber ich werde beide OS neu installiert.

Win11 Revi ist ganz ok ~ aber ich kenne nur die alte Version. (v1702)
Für Win10 wäre Ghost besser.

Ghost 11 ist schlecht und Revi10 war nicht gut 🤭
Stock11 ist eine reine Katastrophe

 

[tibcsi0407]

Why would you only test ~31GB in VST/VT3 when you have 48GB in total?

Because the bitrate is faster with this and hammers the IMC harder. This is not memtest,. It's IMC test. For the Dimms there is the 20 cycles Usmus with 98-99 usage.

 

[Veii]

and hammers the IMC harder. This is not memtest,. It's IMC test

Unsure~
Y-cruncher itself is not a memtest.

But HCI/Karhu & TM5 are different kind of memtests which equally let Ring be utilized.

VT3 messes with:
~ E-Cores
~ Ring
~ L$
~ ^ Which then is offloaded to mem

Its not about what performs faster, but what loads the CPU completely.
14,15,16,17 has a target to fill Backend and MemSubsystem as best as it can.

After inputting curve values , please share a visual curve before you start any tests.
I need to make sure lowest parts are ok, because you put me in a position to guess blind ~ to help you invest less time.

I don't feel like being Tech Support
But so far its ok~~
Need you to find a way to invest the time and work by yourself further, or not bother with it
I got no DDR5 system anymore.

---

CTL0 DQ should be somewhere between (181-99 // 180-98 // 182-100)
Somewhere there, and before you/we play with them ~ we need to test CPU stability

That is SFT and doublecheck if something messed up on VST+VT3.
8533.33333(∞)4 , is such an unfriendly clock.

 

[tibcsi0407]

Unsure~
Y-cruncher itself is not a memtest.

But HCI/Karhu & TM5 are different kind of memtests which equally let Ring be utilized.

VT3 messes with:
~ E Cores
~Ring
~L$
~ ^ Which then is offloaded to mem
Anhang anzeigen 958532
It's not about what performs faster, but what loads the CPU completely.
14,15,16,17 has a target to fill Backend and MemSubsystem as best as it can.

After inputting curve values , please share a visual curve before you start any tests.
I need to make sure lowest parts are ok, because you put me in a position to guess blind ~ to help you invest less time.

I don't feel like being tech support
But so far it's ok~~
Need you to find a way to invest the time and work by yourself further, or not bother with it
I got no DDR5 system anymore.

---

Anhang anzeigen 958535
CTL0 DQ should be somewhere between (181-99 // 180-98 // 182-100)
Somewhere there, and before you/we play with them ~ we need to test CPU stability

That is SFT and double check if something messed up on VST+VT3.
8533.33333(∞)4 , is such an unfriendly clock.

Thank you, I will do that later today.

 

[Neon Knights]

Thank you, I will do that later today.

Hi, i can see you have the Apex Encore, but which general RAM Sticks do you have 8000 , 8200 or 8400?

 

[robonz]

Because the bitrate is faster with this and hammers the IMC harder. This is not memtest,. It's IMC test. For the Dimms there is the 20 cycles Usmus with 98-99 usage.

You are better off „hammering“ under complete load, otherwise the CPU is not working with the entire available memory. TM5 with 1usmus‘ profile tends to error out between 20 and 25 cycles in my experience that’s why I do 25.

Old one with the 13900K:



Moin @Veii

 

[tibcsi0407]

Hi, i can see you have the Apex Encore, but which general RAM Sticks do you have 8000 , 8200 or 8400?

It's Gskill 8000.

Beitrag automatisch zusammengeführt: 12.01.2024

You are better off „hammering“ under complete load, otherwise the CPU is not working with the entire available memory. TM5 with 1usmus‘ profile tends to error out between 20 and 25 cycles in my experience that’s why I do 25.

Old one with the 13900K:

Anhang anzeigen 958548

Moin @Veii

Then I will use higher amount next time. Thank you!

Beitrag automatisch zusammengeführt: 12.01.2024

Ive factored that in

Thank you.
Your values made my computer totally freeze, so tried to do it myself. Probably it's due to the AC_LL 0.35 (my bad.. )
My method was to use the CB die sense table from my "stable" profile and put that values into the AC_LL 0.35 profile.
So with using the suggestedd AC_LL 0.35 I needed smaller offsets than what you suggested.
Here is a video I made, just a small one, this SFT is not a funny load. (here is the one with the screen recorder: video )
A bit shaky, done it by hands. I made one with a capturer too, but that reduced the Y bitrate.
The bitrate on VT3 is much lower than what it was with my previous setting, even though the frequency is the same.
Probably something limits it.

Here is the default AC_LL 0.35 curve and the new one.
There is a drop on some small frequency which should be corrected later. But it's parallel with the default curve.

 

[X909]

At least on ASUS Boards negative VF-offsets do not work. I set a global -0.125V OffSett an partly compensate it with positive offsets in the V/F curve as a workaround.

 

[tibcsi0407]

At least on ASUS Boards negative VF-offsets do not work. I set a global -0.125V OffSett an partly compensate it with positive offsets in the V/F curve as a workaround.

It works for me.
Maybe it fixed on 14th gen.

 

[Veii]

Here is a video I made, just a small one, this SFT is not a funny load. (here is the one with the screen recorder: video )
A bit shaky, done it by hands. I made one with a capturer too, but that reduced the Y bitrate.

Its exactly what i wanted. Thank you

Tibcsi:

SFT , 5.7-4.5

GVID 1214
Vout 1208

UCVID 1304
SAVID 1181
L2$ 1136

IA_PL4 NO
351W, 77° Burst
==================================
VST, 5.7*-4.5
// * slight throttle

GVID 1238
Vout 1234

UCVID,SAVID,L2$ =//=

IA_PL4 NO
330W, 72° Burst
==================================
VT3, 5.7*-4.5
// * slight throttle

GVID 1233
Vout 1225

UCVID,SAVID,L2$ =//=

IA_PL4 Y/NO
RING Hit PL4
330W, 75° Burst

Thank you.
Your values made my computer totally freeze, so tried to do it myself. Probably it's due to the AC_LL 0.35 (my bad.. )
My method was to use the CB die sense table from my "stable" profile and put that values into the AC_LL 0.35 profile.
So with using the suggestedd AC_LL 0.35 I needed smaller offsets than what you suggested.

It is completely my fault for making a calculation mistake between P1-P3
A freeze is not bad
You do have 60mV headroom downwards.
I have to work blind and it was difficult.

Please retry my suggestion at 0.48mΩ AC_LL
That shouldd be plenty to make it to windows.
I need to visually see the curve, because there is a chance that i made a "between steps" mistakes.
Like 1-2mV off.
My math might have been 6-8mV off, but more importantly i need to check transitions & the ~2.5GHz curve.
Sorry 🙇‍♂️ Sadly Peters skillset is still far beyond mine. I'm just an amateur with no talents, except that i learn fast~

At least on ASUS Boards negative VF-offsets do not work.

Highest voltage wins.
Learned on 12th gen, it changed with 13-14th gen.
Basics remain. 12th gen had individual points.
13-14th gen algo is more dynamic. Every point influences the other.

13-14th gen also need to be careful with transition steps.
Please read those notes carefully. Point 3 & 4

EDIT:
V/F is Boardpartner exclusive.
Intel FW Feature.

It's on the Boardpartner's work to implement SVID presets
& give freedom for core/e-core/ring V/F modification access.

 

[zebra_hun]

It's Gskill 8000.

Beitrag automatisch zusammengeführt: 12.01.2024

Then I will use higher amount next time. Thank you!

Beitrag automatisch zusammengeführt: 12.01.2024

Thank you.
Your values made my computer totally freeze, so tried to do it myself. Probably it's due to the AC_LL 0.35 (my bad.. )
My method was to use the CB die sense table from my "stable" profile and put that values into the AC_LL 0.35 profile.
So with using the suggestedd AC_LL 0.35 I needed smaller offsets than what you suggested.
Here is a video I made, just a small one, this SFT is not a funny load. (here is the one with the screen recorder: video )
A bit shaky, done it by hands. I made one with a capturer too, but that reduced the Y bitrate.
The bitrate on VT3 is much lower than what it was with my previous setting, even though the frequency is the same.
Probably something limits it.

Here is the default AC_LL 0.35 curve and the new one.
There is a drop on some small frequency which should be corrected later. But it's parallel with the default curve.
Anhang anzeigen 958561

Good to see, i'm not alone with this work i visit your work now.
We can later maybe via WhatsApp or Viber to talk and show us this online
I can good quality video givin, and to show my settings. In windows works good Team viewer.
Share your setting in our hun Forum, too.

 

[tibcsi0407]

Its exactly what i wanted. Thank you

Tibcsi:

SFT , 5.7-4.5

GVID 1214
Vout 1208

UCVID 1304
SAVID 1181
L2$ 1136

IA_PL4 NO
351W, 77° Burst
==================================
VST, 5.7*-4.5
// * slight throttle

GVID 1238
Vout 1234

UCVID,SAVID,L2$ =//=

IA_PL4 NO
330W, 72° Burst
==================================
VT3, 5.7*-4.5
// * slight throttle

GVID 1233
Vout 1225

UCVID,SAVID,L2$ =//=

IA_PL4 Y/NO
RING Hit PL4
330W, 75° Burst

It is completely my fault for making a calculation mistake between P1-P3
A freeze is not bad
You do have 60mV headroom downwards.
I have to work blind and it was difficult.

Please retry my suggestion at 0.48mΩ AC_LL
That shouldd be plenty to make it to windows.
I need to visually see the curve, because there is a chance that i made a "between steps" mistakes.
Like 1-2mV off.
My math might have been 6-8mV off, but more importantly i need to check transitions & the ~2.5GHz curve.
Sorry 🙇‍♂️ Sadly Peters skillset is still far beyond mine. I'm just an amateur with no talents, except that i learn fast~

Highest voltage wins.
Learned on 12th gen, it changed with 13-14th gen.
Basics remain. 12th gen had individual points.
13-14th gen algo is more dynamic. Every point influences the other.

13-14th gen also need to be careful with transition steps.
Please read those notes carefully. Point 3 & 4
Anhang anzeigen 958581
EDIT:
V/F is Boardpartner exclusive.
Intel FW Feature.

It's on the Boardpartner's work to implement SVID presets
& give freedom for core/e-core/ring V/F modification access.

Hey,

Thank you for the answer!

Updated the settings acc. to your values and AC_LL 0.48
The slope looks much better for my eyes, let's see the SFT.

Beitrag automatisch zusammengeführt: 12.01.2024

Good to see, i'm not alone with this work i visit your work now.
We can later maybe via WhatsApp or Viber to talk and show us this online
I can good quality video givin, and to show my settings. In windows works good Team viewer.
Share your setting in our hun Forum, too.

Of course we can. Just need some fine tuning on this one too.

 

[Brandur]

Bei wem laufen die G.Skill 8400er stable und mit welchen Spannungen?

 

[Veii]

The slope looks much better for my eyes, let's see the SFT.

Wait for my answer pls
I need a visual inspection and as thought
This drops far too low

Sec~

You can sanity check FFT for like 6-7 loops
Give it minimum 15minutes.
Till i get something

I just came back home
Am free for 2-3 hours. DDR4 Shenanigans passed.
I'll try to be a bit (tech) productive today.

Beitrag automatisch zusammengeführt: 12.01.2024

Share your setting in our hun Forum, too.

May you wait slightly.
I feel its not a flawless result.
Would be slightly embarrassing in the current state~

Its just "functional"
Thats not enough~

---

Spoiler: AMD™ ~ AVFS TSMC 7nm

Spoiler: Intel™ ~ 10ESF [10nm]

Goal

But we lack E-Core and ring curve-access.
There is still plenty for work open. Need to work with supply offsets for Ring & E-Cores

 

[tibcsi0407]

Wait for my answer pls
I need a visual inspection and as thought
This drops far too low
Anhang anzeigen 958597
Sec~

You can sanity check FFT for like 6-7 loops
Give it at least 15 minutes.
Till I get something

I just came back home
Am free for 2-3 hours. DDR4 Shenanigans passed.
I'll try to be a bit (tech) productive today.

Beitrag automatisch zusammengeführt: 12.01.2024

Maybe you wait a little.
I feel its not a flawless result.
Would be slightly embarrassing in the current state~

It's just "functional"
That's not enough~

---

Anhang anzeigen 958599
But we lack E-Core an ring curve access.
There is still plenty for work open. Need to work with supply offsets for Ring & E-Cores

I ran a bit forward, but I will leave the PC soon.
I was able to boot with the IA AC 0.48 but it was insta BSOD with SFT. But I slowly raised the AC to 0.68 and I can pass. Load Vcore on SFT is 1.199V
New videos: capture , phone

Voltage is lower, but no more limit reason, which is good.
Bitrates are still far from my previous tests.

 

[Veii]

I ran a bit forward, but I will leave the PC soon.

Can you give on this a scroll capture, phone or merging two screenshots
Of the listed (target) DieSense (VID) for every 100MHz step ?

The written out values, just to be sure

I was able to boot with the IA AC 0.48 but it was insta BSOD with SFT. But I slowly raised the AC to 0.68 and I can pass. Load Vcore on SFT is 1.199V

We target near 1150mV CoreVID or 1160ish Vout.
Around 1180 for VT3.

We hit atm 1230 Vout
But its not just "lower one point"
if you lower one point or transitions are bad, it will crash.
Target is to lower min and mid points, to smooth peak ~ soo overall whole curve drops.
LL(C) Telemetry faking is a linear drop. Substrate has not linear voltage scaling.

I need you later to check bootable high-voltage.
To make sure you have same very-leaky but XOC SA/MC stable chip.

Nonleaky are very different than those 1.46+ ones.
Some are high leakage some are just "failed" bad bins.
Some need high voltage curve due to thermal target, some 14700K rebrands need more voltage or fail target clock
There are many variables why it is how it is.

Important is that ICCMAX is not hit, for any of those harsh loads.
PL4 itself will bother for memOC. Its not just the cores~
// But it has to exist for health, safety and for jitter/transient loadchange spikes
We can fake and bypass, but if we start putting bandaid on bandaid like PLL usage - this leads to nowhere